An Adaptive Neighborhood Ant Colony Optimization Framework for Multiple Uncertain Depots Multiple Travelling Salesmen Problem

초록

Multiple uncertain depots multiple traveling salesmen problem (MUD-MTSP) aims to find optimal routes for multiple salesmen, which involves depot selection optimization, city allocation optimization and route optimization. To solve this problem effectively, this paper devises an adaptive neighborhood ant colony optimization (AN-ACO) framework. First, AN-ACO maintains an ant team containing the same number of ants as that of salesmen to build a feasible solution with each ant constructing a route for one salesman. Second, an adaptive depot selection scheme is designed to adaptively pick one depot for each salesman from all cities with their selection probabilities updated adaptively based on the feedback from the solutions built by ant teams. Third, an adaptive neighborhood candidate city selection scheme is devised for ants to construct their routes efficiently. Fourth, a route length balance biased ant selection strategy is proposed for each ant team to build as balanced routes as possible. Fifth, an adaptive local search method is designated to adaptively execute a newly designed route merging and splitting scheme along with 2-opt on the solutions built by ant teams to refine their quality. At last, to verify the effectiveness and the generality of AN-ACO, this paper instantiates it with five classic ACO algorithms and then conducts experiments on three scales of MUD-MTSP instances on the basis of 18 TSPLIB instances with different numbers of both cities and salesmen. Experimental results substantiate that the five instantiated AN-ACOs significantly outperform 9 latest MTSP algorithms and preserve good scalability and robustness in solving MUD-MTSP. © 1997-2012 IEEE.

키워드